Voice Train

1 · Author voice (Malachi)

Speak and train with my voice — self-recorded samples only; registry locked to Malachi Author. Site / legal name: Malachai Grant. The official sample script uses Malachi.

How to provide WAV / MP3

1. Record a clean reading of the official script below (quiet room, one speaker, no music).
2. Preferred format: mono WAV, ~24 kHz, 16-bit; MP3 also accepted. Aim 90–120 s if you can; shorter is OK.
3. Audio ready: malachi-sample-01.wav / .mp3 (take 2, ~102s) under site/assets/voices/author/. Re-drop only to replace.
4. HF clone live: interim Author preview at malachi-chatterbox-clone.mp3 (ResembleAI/Chatterbox). Listen in the preview section below.
5. xAI still blocked: Custom Voices API returns 403 for this team. When unlocked, clone in the xAI Voice Library, copy Voice ID → paste above or set author.voice_id in the registry and republish. Until then Listen/Talk Author persona falls back to built-in eve.
6. Optional Desktop drop: C:\Users\Malac\Desktop\fulcrum-voices\ for agent import.

Official author sample script

Take 2 on disk (~102s, clear). Listen below with captions + live transcript. Official script remains the clone source of truth. HF Chatterbox preview is separate (interim while xAI is blocked).

Hello, my name is Malachi. This is the test of the universal hierarchy, where I'd explain all of the things that account to the universe and all the things that include everyday life of databases. I'm 19 years of age. I live in Australia, New South Wales, and I am a polymath, a visionary intelligence by heart, an INFJ-OC-AH. I know that sounds very too much, but it is the very soul and very much part of me that makes me flow with energy and the cycle of life, and I will be proud of that moment, and I appreciate that. Thank you.

And with all the things that I do, I really do think that everything that I try is within the best interests of the universe.

And it is not the cockiness that lays within the lies, it is within the truth of the clothes that we reveal ourselves, not through hidden of sin, but to the truth of our own self's consciousness.

Download / open plain text · source of truth for recording.

2 · Commissioned / licensed voices

Registry slots with consent fields: name, license, consent date, sample path, voice_id. Empty until paperwork exists.

3 · Public domain / CC speech

Discoverable sources (Internet Archive, LibriVox, NASA media terms, Freesound CC0, Common Voice). Link-only until a human verifies redistribution — no bulk downloads into the site. Separate from the World Voices embeds gallery (scientists on TED/YouTube — never training rips).

4 · Original beds (Presence spiral)

Procedural or HF-generated atmosphere into site/assets/audio/beds/. Never clone famous people. Spotify remains embed-only.

Generator: site/scripts/generate-presence-beds.py · notes in site/assets/audio/beds/README.md · full checklist in book/VOICE-TRAIN.md.

Docs & hard rules

• No TikTok / YouTube / Twitch rips.
• No training on celebrities without written license in the commissioned registry.
• Author clone = Malachi self-samples only.
• Guide: keep VOICE-TRAIN.md next to the book root for xAI clone steps, optional HF LoRA, and consent checklist.

Audio Station playlist

Author clone batch (Chatterbox). Ready clips play here with matching transcript. Same playlist as Movie hub. Remaining lane/hub clips pending ZeroGPU quota.